基于T2FLAIR和DTI的影像组学模型对脑胶质瘤免疫组化特征的预测研究  

Prediction of immunohistochemistry features of gliomas by radiomics model bases on T2 FLAIR and DT

作　　者：李静[1] 孙静涛[1] 王可欣 张耀峰 李世佳 王俊奇[4] 韩月明 张东香[1] LI Jing;SUN Jingtao;WANG Kerin;ZHANG Yaofeng;LI Shijia;WANG Junqi;HAN Yueming;ZHANG Dongxiang(Debartment of Radiology.Tangshan Women and Children's Hospital,Tangshan 063000,China)

机构地区：[1]河北省唐山市妇幼保健院放射科,063000 [2]首都医科大学基础医学院 [3]北京赛迈特锐医学科技有限公司 [4]唐山市古冶区医院内科

出　　处：《中国煤炭工业医学杂志》2023年第5期497-502,共6页Chinese Journal of Coal Industry Medicine

基　　金：河北省医学科学研究课题(编号:20221754)。

摘　　要：目的探索基于T2 FLAIR和DTI的影像组学模型对脑胶质瘤IHC特征的预测效能。方法回顾性分析2015年3月—2019年9月确诊脑胶质瘤患者中48例术前MRI影像和临床数据用于影像组学模型训练。收集术后病理分级、CD34、Ki-67、p53、S-100、syn、vimentin结果。手工标注T2 FLAIR脑胶质瘤区域,在工作站上测量DTI参数。将患者性别、年龄、肿瘤位置、DTI参数及T2 FLAIR图像联合用于影像组学建模,将数据按7︰3的比例随机分为训练集和测试集,并进行特征提取、降维、选择和分类器训练等步骤,训练影像组学模型。以ROC曲线评价模型效能。结果测试集中脑胶质瘤病理分级、S-100、vimentin、syn、p53、K-i67、CD34的ROC曲线下面积分别为0.910、0.963、0.975、0.950、0.917、0.979和0.933。结论基于T2 FLAIR的影像组学模型预测脑胶质瘤IHC特征有一定的准确性,而基于DTI参数预测IHC是无效的,尚需扩大样本量进一步研究。Objective To explore the predictive efficacy of radiomics model of immunohistochemistry(IHC)features based on T2 fluid attenuated inversion recovery(T2 FLAIR)and diffusion tensor imaging(DTI).Methods This retrospective study collected the patients with glioma from March 2015 to September 2019.The patients underwent MRI examination before surgery.Forty-eight patients were collected who had qualified T2FLAIR and DTI images.The post operation pathology and immunohistology results were collected,inclu-ding grade of the tumor,CD34,Ki-67,p53,S-100,syn and vimentin.The labels of gliomas were manually an-notated by two radiologists on T2 FLAIR images,and the parameters were measured on the DTI parametric maps.The age,gender,location of tumor and DTI parameters were combined with the T2 FLAIR images for the training of the radiomics model.The data were randomly divided into training set and test set in the ratio of 7:3.The radiomics model was trained through the process of feature extraction,feature dimension reduc-tion,feature selection,and classifier training.After obtaining the prediction results of the test set,the model ef-ficacy was evaluated by receiver operating characteristic curve(ROC).Results The AUC in the test set for grade of the tumor,S-100,vimentin,syn,p53,and CD34 was 0.910,0.963,0.975,0.950,0.917,0.979,and O.933,respectively.Conclusion The radiomics model based on the T2 FLAIR images has certain accuracy for predicting IHC of the glioma,but predicting IHC based on DTI parameters is invalid and further research is needed to expand the sample size.

关 键 词：脑胶质瘤 免疫组织化学 扩散张量成像 T2液体衰减反转恢复 影像组学 

分 类 号：R739.4[医药卫生—肿瘤]